Received 20 February 2014; revised 21 March 2014; accepted 28 March 2014

ABSTRACT

Salvia triloba, traditionally known as Greek sage, has been used to enhance memory, as a sedative and to treat headaches. Pharmacological evaluation of purified extracts and isolated compounds of S. triloba were carried out on functional assays using two-electrode voltage clamp methods on recombinant GABA receptors expressed in Xenopus laevis oocytes. Bio-assay guided fractionation led to seven compounds being isolated from S. triloba: ursolic acid, carnosol, oleanolic acid, salvi- genin, rosmanol, cirsimaritin and hispidulin. The purified extracts of S. triloba inhibited 54% of the current produced by 300 μM GABA at a₁b₂g_2L GABA_A receptors. Ursolic acid, carnosol, oleanolic acid and rosmanol also acted as negative allosteric modulators. The flavonoids salvigenin, cirsi- maritin and hispidulin acted as positive modulators when applied in the presence of low concen- trations of GABA but in the presence of high concentrations of GABA acted as negative modulators, demonstrating a biphasic action. These results are consistent with the concept that Salvia triloba may have cognition enhancing properties. In most cases these activities are likely to be occurring via different modulatory sites on GABA_A receptor complexes. It may be that the combination of these activities permits cognition enhancement whilst offering protection from convulsant activi- ty.

Keywords:

GABA_A Receptors; Salvia triloba; Flavonoids; Sedative; Anxiolytics; Chromatography; Allosteric Modulation

1. Introduction

Salvia triloba L. (Lamiaceae), traditionally known as Greek sage, is also known by several other names includ- ing Salvia fruticosa, Salvia libanotica, Salvia lobryana, and Salvia cypria due to a taxonomic confusion over the years. It is a perennial Mediterranean herb and has been used to enhance memory [1] , as a sedative [2] and to treat headaches [3] . Many phytochemicals are known to interact with receptors for the major inhibitory neurotransmitter GABA [4] . Such interaction may underlie some of the pharmacological effects of S. triloba. The constituents of an ethanolic extract of S. triloba have been shown to have moderate affinity to GABA_A benzodiazepine receptor sites [5] . Another study demonstrated that compounds isolated from this plant increased the hypnotic effect of hexobarbital in the rat [6] [7] , suggesting a benzodiazepine like effect.

Ionotropic GABA receptors are ligand gated chloride channels that mediate much of the fast inhibitory neuro- transmission in the brain [8] . GABA_A receptors are antagonised by the convulsant alkaloid bicuculline [9] and modulated by drugs such as benzodiazepines and barbiturates as well by a range of phytochemicals including flavonoids [4] [10] . The less numerous homomeric GABA_C receptors are insensitive to bicuculline and most of the chemicals that modulate GABA_A receptors. The receptor complex is a pentamer composed of a combination of five subunits of these subunits (α1-6, β1-3, γ1-3, δ, ε, π, θ). However, most GABA_A receptors appear to be formed by 2α, 2β subunits in combination with one subunits form another class. The most common subunit combination in the human brain was found to be a₁b₂g₂ [8] [11] .

We have studied the effects of extracts and isolated compounds from S. triloba on functional assays using two-electrode voltage clamp methods on recombinant GABA_A receptors containing a₁b₂g_2L subunits expressed in Xenopus laevis oocytes. Using bio-assay guided fractionation we found seven phytochemicals that modulated these receptors.

2. Materials and Methods

2.1. Plant Materials

Dried Salvia triloba plants were purchased from herbal markets located in Amman. The plants were identified by Prof. Dawud AL-Eisawi (Department of Biological Sciences, Faculty of Science, University of Jordan). Voucher specimens were deposited at the herbarium of the University of Jordan.

2.2. Chemicals, Materials and Instrumentation

All chemicals used were purchased from Aldrich Chemical Co. Ltd (St Louis, MO, USA) unless otherwise stated and were of highest commercially available purity. Solvents were distilled by standard techniques prior to use. Silica gel for column chromatography (CC) was performed on silica gel (Merck silica gel 60H, particle size 5 - 40 μm) and Sephadex LH-20 gel. Routine thin layer chromatography (TLC) was performed on Merck alu- minium backed plates, pre-coated with silica (0.2 mm, 60F254).

The UV-Spectra were recorded on Hitachi U-2000 double beam UV/Vis Spectrophotometer. Mass spectra were carried out on a ThermoFinnigan (Waltham, MA, USA) PolarisQ Ion Trap system using a direct exposure probe. Nuclear magnetic resonance ¹H NMR and ¹³C NMR spectra were recorded on 400 MHz Varian Gemini spectrometer (Palo Alto CA, USA) in DMSO-d₆ (Sigma-Aldrich, USA) with tetramethylsilane as internal standard. Melting points were determined using a Stuart (Stone, Staffordshire, UK) SMP10 melting point apparatus.

2.3. Preparation of the Extracts and Solvent Fractionation

The dried material (5 kg) was ground into fine powder and defatted by extraction with petroleum ether at room temperature for 7 days. After filtration, the remaining materials were then extracted extensively with ethanol for (3 times, 10 days each) at room temperature. The resulting ethanol extract was partitioned between CHCl₃ and H₂O (1:1 v/v). The dried chloroform extract was further partitioned between n-hexane and 10% aqueous metha- nol. The dried methanolic extract was then extracted with ethyl acetate to give the final extract.

2.4. Isolation of Constituents

The final extract (43 g) was chromatographed on a silica gel column chromatography (Φ28 × 8 cm) eluted with a gradient of MeOH/CHCl₃ of increasing polarity resulting in six fractions (SI-SVI). Each fraction was further purified by a combination of column chromatography and preparative thin layer chromatography using suitable solvent systems.

The addition of ethyl acetate to the methanolic extracts of S. triloba yielded an impure solid that was then washed with methanol for further purification and afforded ursolic acid (3-hydroxy-urs-12-en-28-oic acid). Add- ing chloroform to fraction SI lead to the precipitation of impure solid which was then recrystallize by methanol to give carnosol as a white crystals. The TLC of the mother liquor of fraction SI showed a major UV active spot resulting in precipitation of an impure solid which give upon further purification by chloroform salvigenin (5-hydroxy-6,7,4’-trimethoxyflavone). Fraction SII was subjected to column chromatography (Φ31 × 4 cm) and was eluted with hexane/CHCl₃ mixtures of increasing polarity until pure chloroform was used providing three subfractions (SII-1- SII-3). Fraction SII-2 afforded a pure white solid that was then identified as oleanolic acid (3-hydroxy-olean-12-en-28-oic acid). The addition of methanol to fraction SIV yielded an impure solid that produced two major UV active spots on TLC with R_f values of 0.4 and 0.3 (2% MeOH/CHCl₃ as the solvent system). This solid was separated by column chromatography on a Sephadex LH-20 gel (Φ21 × 4 cm) to yield rosmanol (R_f = 0.4) and cirsimaritin (R_f = 0.3) (5,4’-dihydroxy-6,7-dimethoxyflavone). An impure solid precipi- tated upon treating fraction SV with methanol. Further washing with hot acetonitrile of this precipitate gave his- pidulin (5,7,4’-trihydroxy-6-methoxyflavone) as pure yellow solid.

2.5. Physical and Spectroscopic Data of the Isolated Compounds

Ursolic acid (1): light white powder, m.p. 209˚C - 220˚C; ESMS m/z (%): 479.4 (100), 457 (5), 439 (15), 425 (44), 413 (12). ¹H-NMR (DMSO-d6) δ ppm: 5.11 (H, t, J = 3.4 Hz, H-12), 2.99 (H, m, H-3), 2.08 (H, d, J = 11.6 Hz, H-18), 1.02 (3H, s, Me-27), 0.89 (3H, s, Me-23), 0.87 (3H, s, Me-26) 0.84 (H, d, J = 6 Hz, H-30), 0.79 (H, d, J = 6 Hz, H-29), 0.73 (3H, s, Me-24), 0.65 (3H, s, Me-25); ¹³C-NMR (DMSO-d6) δ ppm: 178.7 (C-28), 138.6 (C-13), 125.0 (C-12), 77.3 (C-3), 55.2 (C-5), 52.8 (C-18), 47.5 (C-9), 47.3 (C-17), 42.1 (C-14), 39.5 (C-8), 38.9 (C-4), 38.9 (C-19), 38.8 (C-20), 38.7 (C-1), 37.0 (C-10), 36.8 (C-22), 33.2 (C-7), 30.6 (C-21), 28.7 (C-15), 28.0 (C-23), 27.4 (C-2), 24.3 (C-16), 23.7 (C-27), 23.3 (C-11), 21.5 (C-30), 18.4 (C-6), 17.5 (C-29), 17.4 (C-26), 16.5 (C-24), 15.7 (C-25).

Carnosol (2): white crystals, m.p. 209˚C - 220˚C; UV l_max (MeOH) nm: 209 and 285; ESMS m/z (%): 683.4 (100), 331.0 (19); ¹H-NMR (DMSO-d₆) δ ppm: 6.68 (H, s, H-14), 5.44 (1H, dd, J = 3.8, 1.4 Hz, H-7), 3.21 (1H, sept, J = 6.8, H-15), 2.64 (1H, dd, J = 14, 1.6 Hz, H-1a), 2.43 (1H, ddd, J = 14, 4.6 Hz, H-1β), 2.01 (1H, ddd, H-6a), 1.6 (1H, m, H-6β), 1.59 (1H, dd, J = 10.6, 5.4 Hz, H-5), 1.50 (1H, m, H-2β), 1.42 (1H, dd, J = 13, 1.4 Hz, H-3a), 1.21 (1H, ddd, J = 13.4, 3.2 Hz, H-3β), 1.10 (6H, d, J = 6.8, H-17, H-16), 0.79 (3H, s, H-18), 0.77 (3H, s, H-19); ¹³C-NMR (DMSO-d₆) δ ppm: 175.9 (C-20), 143.7 (C-11), 143.5 (C-12), 134.7 (C-13), 132.0 (C-8), 122.3 (C-9), 111.7 (C-14), 77.4 (C-7), 48.3 (C-10), 45.4 (C-5), 41.0 (C-3), 34.6 (C-4), 31.8 (C-18), 29.7 (C-6), 29.2 (C-1), 26.6 (C-15), 23.2 (C-17), 23.1 (C-16), 19.8 (C-19), 19.0 (C-2).

Oleanolic Acid (3): light white powder; ESMS m/z (%): 456 (100), 431 (15), 391 (8); ¹H-NMR (DMSO-d₆) δ ppm: 5.14 (H, t, H-12), 2.98 (H, t, H-3), 2.78 (H, dd, J = 6.8, 4 Hz, H-18), 1.07 (3H, s, Me-27), 0.87 (3H, s, Me-25), 0.85 (6H, s, Me-30, 24), 0.70 (H, s, H-29), 0.66 (3H, s, Me-26), 0.65 (3H, s, Me-23); ¹³C-NMR (DMSO-d₆) δ ppm: 179.0 (C-28), 144.3 (C-13), 122.0 (C-12), 77.3 (C-3), 55.2 (C-5), 47.5 (C-9), 46.1 (C-19), 45.9 (C-17), 41.8 (C-14), 41.3 (C-18), 38.8 (C-8), 38.5 (C-1), 37.0 (C-10), 33.8 (C-21), 33.3 (C-29), 32.9 (C-22), 32.5 (C-7), 30.8 (C-20), 28.7 (C-23), 27.6 (C-15), 27.4 (C-2), 26.0 (C-27), 23.8 (C-30), 23.3 (C-16), 23.1 (C-11), 18.5 (C-6), 17.3 (C-26), 16.5 (C-24), 15.5 (C-25).

Salvigenin (4): light yellow powder; UV l_max (MeOH) nm: 332 (Band I), 277 (Band II); +NaOMe, 327, (Band I), 278 (Band II); +AlCl₃, 349 (Band I), 300 (Band II); +HCl, 349 (Band I), 299 (Band II); ESMS m/z (%): 350.9 (100), 328.9 (4); ¹H-NMR (DMSO-d₆) δ ppm: 12.86 (1H, s, 5-OH), 8.06 (2H, d, J = 8.8 Hz, H-2’, H-6’), 7.11 (2H, d, J = 8.8 Hz, H-3’, H-5’), 6.95 (1H, s, H-8), 6.93 (1H, s, H-3), 3.92 (3H, s, 7-OMe), 3.85 (3H, s, 4’-OMe), 3.72 (3H, s, 6-OMe); ¹³C-NMR (DMSO-d₆) δ ppm: 182.7 (C-4), 164.1 (C-2), 162.9 (C-4’), 159.2 (C-7), 153.1 (C-5), 152.5 (C-9), 132.4 (C-6), 128.8 (C-2’, C-6’), 123.2 (C-1’), 115.1 (C-3’, C-5’), 105.6 (C-10), 103.8 (C-3), 92.1 (C-8), 60.5 (6-OMe), 56.9 (7-OMe), 50.1 (4’-OMe).

Rosmanol (5): light white powder; ESMS m/z (%): 314.9 [M-CH3]⁺ (100), 369.0 (93), 301.0 (28), 347 (8); ¹H-NMR (DMSO-d₆) δ ppm: 6.68 (H, s, H-14), 5.08 (1H, d, J = 4.4 Hz, H-7), 4.13 (1H, t, J = 4 Hz, H-6), 3.21 (1H, sept, J = 1.6 Hz, H-15), 2.59 (1H, d, J = 14.4 Hz, H-1a), 2.57 (1H, m, H-5), 2.53 (1H, ddd, J = 13.8, 4.4 Hz, H-1β), 1.60 (1H, m, H-2a), 1.19 (1H, d, J = 4 Hz, H-3a), 1.49 (1H, m, H-2β), 1.21 (1H, ddd, J = 13.4, 3 Hz, H-3β), 1.25 (3H, d, J = 1.6 Hz, H-17), 1.11 (3H, d, J = 1.6 Hz, H-16), 0.94 (3H, s, H-18), 0.79 (3H, s, H-19); ¹³C-NMR (DMSO-d₆) δ ppm: 175.2 (C-20), 143.9 (C-11), 143.4 (C-12), 134.6 (C-13), 128.8 (C-8), 122.2 (C-9), 115.2 (C-14), 80.0 (C-6), 67.7 (C-7), 55.0 (C-5), 47.9 (C-10), 41.1 (C-3), 34.3 (C-4), 32.3 (C-18), 29.2 (C-1), 26.7 (C-15), 23.3 (C-17), 23.0 (C-16), 21.3 (C-19), 18.9 (C-2).

Cirsimaritin (6): light yellow crystals, m.p. 267˚C - 269˚C; UV l_max (MeOH) nm: 333 (Band I), 275 (Band II); +NaOMe, 387, (Band I), 272 (Band II); +AlCl₃, 353 (Band I), 299 (Band II); +HCl, 353 (Band I), 299 (Band II); ESMS m/z (%): 336.8 (100), 314.9 (43), 300.9 (4); ¹H-NMR (DMSO-d₆) δ ppm: 12.91 (1H, s, 5-OH), 7.94 (2H, d, J = 8.8 Hz, H-2’, H-6’), 6.91 (2H, d, J = 8.8 Hz, H-3’, H-5’), 6.91 (1H, s, H-8), 6.82 (1H, s, H-3), 3.90 (3H, s, 7-OMe), 3.71 (3H, s, 6-OMe); ¹³C-NMR (DMSO-d₆) δ ppm: 182.7(C-4), 164.5 (C-2), 161.9 (C-4’), 159.0 (C-7), 153.1 (C-5), 152.5 (C-9), 132.3 (C-6), 129.0 (C-2’, C-6’), 121.4 (C-1’), 115.4 (C-3’, C-5’), 105.5 (C-10), 103.1 (C-3), 92.0 (C-8), 60.5 (6-OMe), 56.9 (7-OMe).

Hispidulin (7): light yellow powder, m.p. 290˚C - 292˚C; UV l_max (MeOH) nm: 334 (Band I), 274 (Band II); +NaOMe, 392, (Band I), 276 (Band II); +AlCl₃, 353 (Band I), 300 (Band II); +HCl, 351 (Band I), 299 (Band II); ESMS m/z (%): 322.9 (100), 300.9 (98), 271.0 (13), 238.9 (8), 197.7 (7); ¹H-NMR (DMSO-d₆) δ ppm: 13.06 (1H, s, 5-OH), 7.90 (2H, d, J = 8.8 Hz, H-2’, H-6’), 6.90 (2H, d, J = 8.8 Hz, H-3’, H-5’), 6.76 (1H, s, H-3), 6.57 (1H, s, H-8), 3.73 (3H, s, 6-OMe); ¹³C-NMR (DMSO-d₆) δ ppm: 182.6 (C-4), 164.2 (C-2), 161.6 (C-4’), 157.7 (C-7), 153.2 (C-5), 152.8 (C-9), 131.8 (C-6), 128.9 (C-2’, C-6’), 121.7 (C-1’), 116.4 (C-3’, C-5’), 104.5 (C-10), 102.8 (C-3), 94.7 (C-8), 60.4 (6-OMe).

2.6. Pharmacological Analysis

Pharmacological evaluation of the final extract and the isolated compounds from Salvia triloba were carried out on functional assays using two-electrode voltage clamp methods on recombinant GABA receptors expressed in Xenopus laevis oocytes using the methods described previously [12] . Bio-assay guided fractionation led to seven compounds being isolated from S. triloba; these are: ursolic acid, carnosol, oleanolic acid, salvigenin, rosmanol, cirsimaritin and hispidulin.

3. Results

The final extract of Salvia triloba tested at 200 mg/mL did not show any effect at uninjected oocytes. The ex- tract of S. triloba activated the receptor by 53.2% ± 0.2% when applied alone at a₁b₂g_2L GABA receptors and in- hibited currents due to 300 μM GABA by 54.1% ± 0.2%.

3.1. Ursolic Acid

Ursolic acid it did not show any effect at sham injected oocytes or at a₁b₂g_2L GABA_A receptors when adminis- tered alone. However, At 300 µM ursolic acid inhibited currents due to 100 µM GABA by 29.0% ± 0.4% (Fig- ure 1, Table 1). GABA dose response curves were carried out both with and without ursolic acid (Figure 1) at a₁b₂g_2L GABA_A receptors. Ursolic acid (100 µM) shifted the GABA dose response curves at a₁b₂g_2L GABA_A receptors to the right, increasing the mean GABA EC₅₀ from 123.3 to 210.4 µM (95% CI: 29.4 to 1508), with a Hill slope of 0.48 ± 0.16 (compared to 0.73 ± 0.07 in the case of GABA alone). The effect of ursolic acid was greater at higher concentrations of GABA, with no effect at concentrations below 30 µM GABA.

3.2. Oleanolic Acid

Oleanolic acid (Figure 2) did not have any effect on sham-injected oocytes (n = 3, data not shown). Also, olea- nolic acid demonstrated no activity at a₁b₂g_2L GABA receptors when administered alone but inhibited currents due to 3000 µM GABA by 31.0% ± 1.6% (Table 1).

3.3. Carnosol

Carnosol (Figure 3) did not have any effect at sham-injected oocytes (n = 3, data not shown) or at a₁b₂g_2L GABA_A receptors when administered alone, but inhibited currents due to 100 µM GABA with an IC₅₀ of 80.11 µM (95% CI: 69.6 to 92.3) and a Hill coefficient of 6.0 ± 1.5 (Figure 3(a)). GABA dose response curves were

^aPercentage inhibition of maximal GABA response by 100 µM compound, ^bPercentage enhancement of GABA EC₅response by 100 µM compound.

in the presence and absence of carnosol (100, 300 and 1000 µM) at a₁b₂g_2L GABA receptors. Carnosol had no significant effect on EC₅₀ of GABA. However, 100, 300 and 1000 µM of carnosol inhibited currents due to 3000 µM GABA by 17.5% ± 4.0%, 40.7% ± 3.3% and 58.5% ± 3.5%, respectively (Figure 3(b)). Carnosol appears to have a slight positive modulation of GABA at lower does of GABA.

3.4. Salvigenin

Salvigenin (Figure 4) did not have any effect on sham-injected oocytes and had no activity at a₁b₂g_2L or a₁b₂ GABA_A receptors when administered alone. Salvigenin exhibited a biphasic effect at a₁b₂g_2L GABA_A receptors. At high concentrations of GABA (100 µM), salvigenin (100 µM) inhibited currents due to GABA by 40.2% ± 0.7% and this action was unaffected by the addition of 10 µM flumazenil, while at low concentrations of GABA (10 µM) salvigenin positively modulated the current by 89.9% ± 1.5% (Table 1). This enhancement was re- duced to 21.8% ± 0.2% by flumazenil (10 μM). At a₁b₂ GABA_A receptors (data not shown), no potentiation of current due to low doses of GABA was observed, however salvigenin reduced the current due to 10 µM GABA by 49.2% ± 0.3% at 100 µM.

3.5. Rosmanol

Rosmanol (Figure 5) did not have any effect on sham-injected oocytes (n = 3, data not shown) and had no activ- ity at a₁b₂g_2L or a₁b₂ GABA_A receptors when administered alone. Rosmanol appears to have two modes of ac- tion at a₁b₂g_2L GABA_A receptors. At high concentrations of GABA (3000 µM), rosmanol (100 µM) inhibited currents due to GABA by 26.7% ± 2.1% and this inhibition was not affected by the addition of flumazenil (10 µM), while at low concentrations of GABA (10 µM) rosmanol positively modulated the current by 99.7% ± 0.5% (Table 1) and this enhancement was reduced to 50.8% ± 0.3% by flumazenil (10 μM). At a₁b₂ GABA_A recep- tors, no enhancement of the GABA response was observed, however, rosmanol inhibited currents due to 10 µM GABA by 29.2% ± 0.2% at 100 µM (data not shown), similar to its effect at a₁b₂g_2L GABA_A receptors. Ros- manol (100 µM) decreased the maximal GABA response by approximately 50% and shifted the GABA EC₅₀ and decreased the EC₅₀ of GABA from 123.3 (95% CI: 91.5 to 166.2) to 19.51 µM (95% CI: 9.5 to 40.1), with a Hill slope of 0.88 ± 0.15 (Figure 5).

3.6. Cirsimaritin

Cirsimaritin (Figure 6) exhibited a biphasic effect at a₁b₂g_2L receptors, inhibiting currents due to 100 µM GABA by 23.0% ± 0.5% at 100 µM and positively modulating currents due to 10 µM GABA by 89.9% ± 1.5%. The inhibitory effect of cirsimaritin at high concentrations of GABA showed further inhibition in the presence of 10 µM flumazenil, however, this is most likely due to receptor desensitization. Flumazenil had no effect on the enhancement of the GABA response at low concentrations of GABA. Cirsimaritin was inactive at a₁b₂ GABA_A receptors (data not shown). At a₁b₂g_2L GABA_A receptors cirsimaritin (60 µM) shifted the GABA dose response curve to the right at low GABA concentrations (<60 µM) and to the left at higher GABA concentra- tions.

3.7. Hispidulin

Hispidulin (Figure 7) produced no effect on sham-injected oocytes (n = 3, data not shown) or at a₁b₂g_2L and a₁b₂GABA_A receptors when administered alone but inhibited currents due to 100 µM GABA with an IC₅₀ of 81.7 µM (95% CI: 45.24 to 147.8) and a Hill coefficient of 1.6 ± 0.64. At a₁b₂g_2L GABA_A receptorshispidulin (100 µM) shifted the GABA dose response curve to the right at GABA concentrations less than 300 µM and to the left at higher GABA concentrations. This resulted in a change to the GABA EC₅₀ from 123 µM (95% CI: 91.5 to 166.2) to 60 µM (95% CI: 19.5 to 185.5). Hispidulin acts as a positive modulator when applied at low concentrations of GABA but at high concentrations it acts as a negative modulator. This resulted in a change to the GABA EC₅₀ from 123 µM (95% CI: 91.5 to 166.2) to 60 µM (95% CI: 19.5 to 185.5). Hispidulin acts as a

positive modulator when applied at low concentrations of GABA but at high concentrations it acts as a negative modulator. The effect of hispidulin on low concentrations of GABA was blocked by 10 µM flumazenil. Howev- er, the effect of hispidulin on high concentrations of GABA was not affected by 10 µM flumazenil. At a₁b₂ GABA_A receptors, hispidulin inhibited currents due to 100 µM GABA by 52.2% ± 0.5% at 100 µM (Table 1). The maximum concentration of hispidulin applied was 300 µM due to the limits of its solubility.

4. Discussion

Ursolic acid is a pentacyclic triterpenoid that exists widely in many medicinal plants used in traditional medicine, and it appears that ursolic acid and carnosol are partly responsible for the antitumorigenic activity of rosemary [13] . Although ursolic acid has been reported to inhibit GABA transaminase (GABA-T) [14] , to date, no stu- dies have investigated the action of ursolic acid at GABA_A receptors. In the current study ursolic acid inhibited currents due to 100 µM GABA with an IC₅₀ of 98.7 µM and shifted the GABA dose response curve at a₁b₂g_2L GABA_A receptors to the right increasing the GABA EC₅₀approximately two-fold to 210.4 µM. Interestingly, this result is in contrast to the in vivo results reported in a recent study by Taviano et al. in which ursolic acid showed significant CNS depressant properties in mice when administrated orally. In this study ursolic acid pro- duced a potentiation of pentobarbital-induced sleeping time and a protective action against pentylenetetrazol (PTZ) induced convulsion [15] . As GABA is an inhibitory neurotransmitter, inhibition of the GABA response would be expected to have an overall excitatory response, and in fact most GABA_A antagonists are convulsants. This suggests that ursolic acid is exerting its CNS depressant actions either through a different mechanism or a different GABA_A receptor sub-type.

Carnosol is primarily responsible for the high antioxidant activity of Rosmarinus officinalis [13] [16] [17] . Carnosol significantly increased tyrosine hydroxylase activity suggesting that it may be a potential for the treat- ment for Parkinson’s disease (PD) [18] . In a study by Rutherford and colleagues, carnosol inhibited the binding of t-butylphosphorothionate (TBPS) to the chloride channel of the GABA/benzodiazepine receptor but it had no effect on the binding of muscimol, diazepam or flunitrazepam, suggesting that the site of action of carnosol to be directly on the chloride channel [19] .

The current study investigates the effects of carnosol at human recombinant a₁b₂g_2L receptors expressed in Xenopus laevis oocytes. Carnosol had no activity at a₁b₂g_2L GABA receptors when administered alone but inhi- bited currents due to 100 µM GABA with an IC₅₀ of 80.11 µM. Carnosol was found to non-competitively block GABA, being more effective at high dose of GABA. These features are characteristic of a channel blocker.

Oleanolic acid is an isomer of ursolic acid, differing only on the location of one methyl group. In a study by Ha et al., oleanolic acid was tested for its ability to modulate binding to GABA_A benzodiazepine receptors in vi- tro; in this study oleanolic acid did not affect the binding of [³H] Ro15-1788 or [³H] flunitrazepam in the pres- ence of GABA [20] . In another study oleanolic acid was found to inhibit GABA transaminase (GABA-T) by 20% at 10 μM/mL [14] . In the current study oleanolic acid inhibited currents due to 3000 µM GABA by 30% most probably by acting as a channel blocker with weaker activity than carnosol.

To date, no studies have investigated the action of salvigenin at GABA_A receptors. In the current study salvi- genin inhibited currents due to 100 µM GABA by 40% at 100 µM and positively modulated currents due to 10 µM GABA by 90%. This compound appears to have two modes of action, with positive modulation occurring via the high-affinity flumazenil benzodiazepine site and the inhibitory action occurring via a site that is inde- pendent of the presence of the g-subunit.

Rosmanol is one of the compounds that gives sage free radical scavenging activity [21] [22] . 7-Methoxyros- manol and galdosol are two derivatives of rosmanol that have been shown to competitively inhibit³H-flumazenil bindingto the benzodiazepine receptor with IC₅₀ values of 7.2 and 0.8 µM, respectively [23] . In the current stu- dyrosmanol at 100 µM inhibited currents due to 3000 µM GABA by 25%. GABA dose response curves were carried out both without and with rosmanol (100 µM) at a₁b₂g_2L GABA receptors shifting the GABA dose re- sponse curves to the right with a GABA EC₅₀ 19.51 µM. Rosmanol has a biphasic mode of action, positively modulating the effect of GABA at low concentrations of GABA and inhibiting the response to high concentra- tions of GABA. This suggests that rosmanol may have two sites of action on the GABA receptor complex.

The two phases of rosmanol at a₁b₂g₂ GABA_A receptors are thought to be mediated via two distinct mechan- isms. The positive modulation of the GABA response by rosmanol at a₁b₂g_2L GABA_A receptors was sensitive to antagonism by flumazenil only at low concentrations of GABA indicating the involvement the “high affinity” benzodiazepine binding site which requires the g subunit. The inhibitory second phase action of rosmanol is not affected by flumazenil and is observed at receptor combinations both with and without a g subunit, indicating that the presence of the g subunit is not a requirement. Therefore the inhibitory phase is not mediated via the high-affinity flumazenil sensitive benzodiazepine site. This inhibitory phase may occur via the “low-affinity” benzodiazepine binding site, which is known to be present on ab combinations or at a novel site independent of benzodiazepine activity.

Competitive inhibition of ³H-flumazenil binding was detected for cirsimaritin when tested for its affinity to the benzodiazepine receptor in a membrane preparation from human frontal cortex with an IC₅₀ of 350 mM [23] . Cirsimaritin inhibited the binding of [methyl-³H] diazepam to rat brain benzodiazepine receptors with an IC₅₀ of 23 µM and induced a small increase in [³⁵S]TBPS binding [24] . In the current study, the positive modulation by cirsimaritin of low concentrations of GABA at a₁b₂g_2L GABA_A receptors was found to be insensitive to anta- gonism by flumazenil indicating that the high affinity benzodiazepine binding site is not involved which sug- gests the involvement of the “low-affinity” binding site, or an alternative novel biding site. At high concentra- tions of GABA, the inhibitory effect of cirsimaritin appeared to be slightly increased by the presence of fluma- zenil, however this is most likely due to desensitization of the GABA receptors. Interestingly, although cirsima- ritin was insensitive to flumazenil, it showed no activity at a₁b₂ GABA_A receptors, indicating that the presence of the g subunit is essential for activity, suggesting that cirsimaritin may modulate the response to GABA via a novel binding site.

The fact that cirsimaritin displaces benzodiazepine binding, yet is flumazenil insensitive in functional assays is not without precedent. For example 6-methylflavone has been reported to competitively displace [³H]-Ro 15-1788 binding in assays on rat brain membranes in vitro and human recombinant GABA_A/BZD receptors ex- pressed in Sf-9 insect cells [25] . However, 6-methylflavone has been shown to be a flumazenil insensitive posi- tive modulator at a₁b₂g_2L GABA_A receptors expressed in Xenopus oocytes [26] . Similarly, amentoflavone has been shown to be on of the most potent non-nitrogen containing ligands for the benzodiazepine site in binding assays with K_i = 6 nM [27] . However, at a₁b₂g_2L GABA_A receptors expressed in Xenopus oocytes, amentofla- vone has been shown to be a flumazenil insensitive negative modulator [28] [29] .

Hispidulinwas tested for its affinity to the benzodiazepine receptor and was found to competitively inhibit the binding of ³H-flumazenil with an IC₅₀’s of 1.3 mM [23] and 8 µM [30] . Synthetic hispidulin was investigated at recombinant GABA_A receptors expressed in Xenopus laevis oocytes and found to act as a positive allosteric modulator at GABA_A receptor subtypes (a_1-3,5,6b₂g₂) being more potent at a_1,2,5b₂g₂ subtypes than at a_3,6b₂g₂ [31] . In the same study, hispidulin was shown also to have an anticonvulsant action in seizure prone mongolian ger- bils and to cross the blood brain barrier. Unlike diazepam, hispidulin was found to act as a positive modulator at a₆b₂g_2L GABA_A receptors at which 10 mM of hispidulin enhanced the action of GABA at these receptors by 65%, with the enhanced response being reduced to 37% by 1 μM flumazenil [31] .

Hispidulin also appears to have a biphasic mode of action at a₁b₂g_2L GABA_A receptors: acting as a positive modulator when applied with low concentrations of GABA but at high concentrations of GABA it acts as a neg- ative modulator. The two phases of hispidulin at a₁b₂g₂ GABA_A appear to be mediated via two distinct mechan- isms. The positive modulation of low concentrations of GABA by hispidulin at a₁b₂g_2L GABA_A receptors was sensitive to antagonism by flumazenil indicating the involvement the “high affinity” benzodiazepine binding site that requires the g subunit. The second phase or inhibitory action of hispidulin is not affected by flumazenil and is observed at receptor combinations both with and without a g subunit, indicating that it is not acting at the high-affinity benzodiazepine binding site.

5. Conclusions

Although previous studies have investigated some of the chemical constituents of sage at the GABA_A/benzo- diazepine complex and cholinergic receptors, few studies have investigated the constituents of Salvia at GABA_A receptors using a functional electrophysiological assay.

GABA_A receptors are known to be implicated in memory and cognition with enhancers of GABA function such as benzodiazepines having a well-documented detrimental effect on cognitive function. However, many GABA inhibitors or antagonists are not sub-type selective, acting at GABA receptors throughout the central nervous system and usually have the significant activity as convulsants, making them unsuitable for use as cognitive enhancers. In this study, the majority of compounds demonstrated some inhibitory activity at a₁b₂g_2L GABA_A receptors, which supports the concept that this herb may have cognition enhancing properties. Interestingly, many of the compounds demonstrated a biphasic activity at a₁b₂g_2L GABA_A receptors, enhancing the activity of GABA at lower concentrations of GABA and showing inhibition at higher GABA concentrations. In most cases these activities are likely to be occurring via different modulatory sites on the GABA_A receptor complex. It may be that the combination of these activities permits cognition enhancement whilst offering protection from convulsant activity.

The action of salvigenin, rosmanol, and hispidulin at low concentrations of GABA were found to be flumaze- nil sensitive, these compounds also demonstrated flumazenil insensitive actions at high concentrations of GABA. Particularly interesting is the action of cirsimaritin, which was found to be flumazenil insensitive, but requires the presence of the g-subunit, suggesting that cirsimaritin may act at a novel modulatory site on the GABA_A re- ceptor complex. Further studies to determine the properties of this site may provide another target for therapeu- tic modulation of GABA_A function.

Conflict of Interest

The authors have no conflict of interest.

Acknowledgements

A. A. gratefully acknowledges the financial support of the Government of Australia for an Endeavour Interna- tional Postgraduate Research Scholarship (EIPRS).

References

NOTES

^*Corresponding author.